On Local Convergence of Stochastic Global Optimization Algorithms
The use of population type nature inspired global optimization algorithms is quite popular due to their attractive natural interpretation and the easy availability of codes. The guarantee of convergence is however doubtful. What this paper investigates is the question how some popular algorithms behave on converging "in the end" to a local minimum point varying dimension and condition of the related Hessian.
The most counterintuitive result is that some popular codes actually do very bad for simple problems.
Basically, this paper calls for investigating algorithms in a systematic way finding out for which cases algorithms do yes or no work well.
The most counterintuitive result is that some popular codes actually do very bad for simple problems.
Basically, this paper calls for investigating algorithms in a systematic way finding out for which cases algorithms do yes or no work well.
Type:
Scientific Paper
Area:
Optimization
Target Group:
Basic
DOI:
10.1007/978-3-030-86976-2_31
Cite as:
Hendrix, E.M.T. and Rocha, A.M.A.C. (2021), On Local Convergence of Stochastic Global Optimization Algorithms, O. Gervasi et al. (Eds.): ICCSA Cagliari, Sept. 13-16 2021, LNCS 12953, Springer, Cham, pp. 456-472, doi: 10.1007/978-3-030-86976-2_31
Author of the review:
Eligius Hendrix
University of Malaga
You have to login to leave a comment. If you are not registered click here